The course

15 chapters, in the order you'd build a subscription business. Each lesson pulls together what every guest said on it. Zoom from a one-line gist to every number.

40 tips with numbers behind them

The single most useful thing each guest said, with the result that backs it.

1. A/B test your plan layout against the traditional two-plan paywall

   The usual paywall shows yearly and monthly upfront, but that layout is not necessarily the best. Mojo tested a single-plan view against it and measured the share of yearly subscribers. Test the default plan presentation rather than assuming the convention works.

   The single-plan variant increased the proportion of yearly subscriptions by 15-20%.

   Mojo · abtest.design

2. Stop paying view bonuses at roughly one million views

   Usually cap bonuses at a million plus. Creators rarely feel a difference between 1 million and 2 million views even though it is twice as many. A video that hits a million often grows to 1.3 to 5 million views, and not paying for that growth is pure margin.

   A video at 1M views will probably grow to 1.3 to 5 million views.

   Jake, Cal AI · App Masters · 4:55

3. Pay affiliates a flat one-time amount per paid signup

   Cal AI's public affiliate program pays $10 for every signup that results in a sale, one time. Aligning incentives makes it a win-win, and a flat one-time payout avoids long-running revenue obligations that could complicate selling the app.

   The $10 payout is roughly 33% of a $30 sale.

   Jake Castillo, Cal AI · Sub Club · 1:06:27

4. Plan on a rolling six months and push for 50 to 100 day milestones

   Plan around the window in which labs release new models, roughly six months, and keep a multi-year vision rather than a multi-year roadmap. The best labs teams go from idea to a meaningful milestone in 50 to 100 days. Remove blockers so people can build.

   Gemini works on a six-month roadmap, and a project that began 48 hours before the interview could not have been planned.

   Josh Woodward, Google Labs · Lenny's Podcast · 43:10

5. Repost any video with 10,000 to 20,000 views and 3% engagement

   When a video reaches 10,000 to 20,000 views with at least 3% engagement, tell the creator to double down. Repost the same video, or redo the same script in the same setting with a different text hook. Reposting the exact video a few more times also works. The videos are often as good as the 10 million view ones; they just missed the right moment in the algorithm.

   If those reposts keep getting at least 10,000 views over the next four to five days, it usually leads to a 500,000 to 1 million view video.

   Dawson Gibbs, Evo Marketing · Superwall · 43:19

6. Run the MVP as an A/B test to prove causation

   DashPass was put behind an A/B test so the team could isolate that more orders came from the program and not from selection bias. Executives had worried that heavy users were simply switching over. The test showed the lift was real, and it kept compounding instead of flattening.

   Order frequency lifted almost 2% in the first 2 or 3 weeks, versus typical tests of about 0.1%, and the lift kept growing from 0.5 to 1% up to 2%.

   Jimmy Liu, DoorDash · Subversive · 19:50

7. Storonsky applied for a UK banking license after reaching 7 million customers and waited almost 5 years

   Revolut began with light e-money and payment licenses because they took about 1 year versus 3 to 4 years for a banking license. Once it had millions of customers, the regulator saw it as far riskier than a zero-customer applicant, since blame would land on them if something went wrong. Storonsky calls this his biggest mistake and says a bank license is needed to win the trust of salaries, wealth and trading.

   The UK license took almost 5 years, and no company with 7 or 8 million UK customers had applied for one before.

   Nik Storonsky, Revolut, Revolut · David Senra · 9:00

8. Bob Nardelli applied GE-style systems to Home Depot and gutted the service culture that made it special

   He replaced knowledgeable floor staff with part-time general retail labor, cut associates per store from 200 to 170, and preferred college-degree store managers, which closed the promotion path. Six sigma treats line employees as replaceable, but at Home Depot a 25-cent washer sale could turn into a $100,000 kitchen. Customer satisfaction fell from near the top to dead last among major US retailers.

   He doubled revenue by doubling stores, while the stock fell 12% over five years against Lowe's up 173%.

   Bernie Marcus, Arthur Blank, Ken Langone and Pat Farah, The Home Depot, The Home Depot · Acquired · 2:09:50

9. Ross Perot said giving Steve Jobs too much money was his mistake at NeXT

   Perot put $20 million into NeXT for 16% in 1986, and Jobs also raised $100 million from Canon and $60 million from IBM. Flush with cash, NeXT bought floating staircases and $10,000 chairs before it had shipped a product. Perot concluded that abundant money removed the hunger a startup needs.

   NeXT burned $39 million in a year when it planned to lose $3 million, and without IBM and Perot's money it would have been dead long before.

   Steve Jobs, NeXT, NeXT · Founders · 35:26

10. Expect pricing page redesigns to lift both free and paid signups

    GitHub's redesign did not just help paid conversion. It raised signups for the free tier and the paid tiers together, so a clearer pricing page can improve the whole funnel.

    20+% increase in free and paid signups.

    Github · abtest.design

11. Hit funnel conversion targets before spending on Meta ads

    Before launching, aim for 5 to 10% install-to-trial activation, over 30% trial-to-paid, and at least 2% install-to-paid. With a hard paywall, aim for 10% or more trial activation. Put the trial on the plan with the highest LTV, either yearly or weekly. Meta amplifies the funnel you have, so a leaky one just burns money.

    1,000 installs at 7% trial activation gives 70 trials, and 30% trial-to-paid gives 21 subscribers, about 2.1% install-to-paid.

    Steve P. Young, App Masters · App Masters · 0:29

12. Use humorous viral content alongside paid ads, but expect lower intent

    A TikTok trend of people crying at low scores from the app brought BoldVoice into feeds in a light-hearted way without ad boosting. Humor drives big spikes but the viewers have less intent than people from paid ads or word of mouth, so treat it as a companion to paid marketing.

    One creator reached about 300 million views in a few months, unboosted by ads.

    Anada Lakra, BoldVoice · Sub Club · 47:30

13. Protect the main tab and say no to extra sidebar entries

    When folding successful experiments into a product, expect a power law in surface usage. Make the main experience great and use it as an entry point to others. Half the job is saying no to another sidebar item or tab, while rolling things in once proven and not killing them too early.

    At Instagram the main feed was about 80% of use, and improving explore might move it from 10 to 15%.

    Ami Vora and Mike Krieger (hosted by Dan Shipper), Anthropic · Lenny's Podcast · 26:48

14. Add Instagram and Facebook Reels to your TikTok-first program

    Instagram has recently outperformed TikTok, while TikTok shadow bans accounts in monetization niches like crypto, trading and income apps. Facebook Reels crossposting is easy, so creators will often post there for a good bonus structure and little or no base increase.

    Instagram is driving probably 10 to 20% more views than TikTok in aggregate over the past 3 to 6 months.

    Canyon (Kenyon Prasad), Sideshift · Superwall · 19:40

15. Track how your user base shifts and respond when the market moves

    Zero measured how many users were on or interested in GLP-1s. The share on a GLP-1 rose from about 5% in early 2025 to a third by May 2026, with another third interested or seeking one. That data convinced the company to add a GLP-1 offering rather than ignore the drug.

    About 5% of users on a GLP-1 at the start of 2025 versus a third by May 2026, plus another third interested.

    Mike Maser, Zero · Subversive · 34:54

16. Ari Emanuel's blow-up in the first merger meeting delayed the William Morris deal four years

    Ari says he screwed up the 2003 meeting badly while Patrick was perfect, and that going crazy does not always work. The tantrum pushed the merger out until the rival was in disarray. He calls it a horrible game of chess.

    The merger that could have happened in 2003 was done around 2009.

    Ari Emanuel, Endeavor, WME Group and TKO, Endeavor / TKO · David Senra · 24:50

17. Michael Eisner rooted against Pixar's Nemo and leaked memo ended his partnership with Steve Jobs

    After seeing an early story reel, Eisner told the board that Nemo would flop and that failure would give Disney leverage in renegotiation. Early reels are always rough in Pixar's process, and the memo leaked. Jobs ended negotiations in January 2004.

    Nemo grossed $871 million in theaters, and about 65 million DVDs were sold for roughly $2 billion.

    Michael Eisner, Frank Wells and Bob Iger, The Walt Disney Company, The Walt Disney Company · Acquired · 2:27:50

18. Phil Knight's biggest regret was missing his sons' childhood while building Nike

    Knight repeatedly vowed to spend more time with Matthew and Travis, kept the promise briefly and then fell back into his routine. He said he gave himself low marks as a father. In his closing reflections, he named this his top regret, one that clashed with his wish to relive the whole Nike story.

    His son Matthew died in a drowning accident in his 20s, as the host notes.

    Phil Knight, Nike, Nike · Founders · 1:01:13

19. Benchmark: personalized plan recommendations can lift upgrades by 10-20%

    Use this range as a reference when testing personalization in onboarding and paywall flows. It comes from Grammarly's test of questionnaire-driven plan suggestions.

    Result: 10-20% increase in upgrade rates.

    Grammarly · abtest.design

20. Remember small uplifts add up to real downloads

    Even a modest conversion lift is worth keeping because it compounds on existing traffic with no extra spend. Steve notes that a 15% lift on 10,000 downloads means 1,500 more downloads without doing anything else.

    15% on 10,000 downloads is an additional 1,500 downloads.

    Steve P. Young, App Masters · App Masters · 14:40

21. Track marketing cash ROI and keep it positive when scaling spend

    Visible's single acquisition metric is marketing cash ROI, essentially ROAS: put a dollar in, get a dollar out that day. It depends on the blend of annual and monthly plans, how much you spend and how fast you scale it. Roughly 50% of acquisition comes via paid, the rest through word of mouth.

    The business stayed cash flow positive through the whole growth journey, reaching about $7M revenue before its Series A.

    Luke Martin-Fuller, Visible · Sub Club · 43:00

22. Judge early startups by operating speed between board meetings

    A top signal is tempo: problems found at one board meeting are root-caused, shipped and measured by the next. Ask how fast the team ships against how long the category normally takes. Speed compounds.

    Ramp was close to shipping cards in about 3 months when it usually takes 9 to 12 months at best, which led Rabois to preempt the Series A. Roelof Botha said Square's pace was the best since PayPal.

    Keith Rabois, Khosla Ventures · Lenny's Podcast · 1:07:00

23. Plan for clones to earn about 20% of the original's views

    The median cloned post gets about 20% of the original's views, so post the same format five to 10 more times to match the original's volume. Some clones do far better, so keep varying them.

    Median clone gets 20% of the original; extreme cases reach two to six times the original.

    Matt Espinoza (host: Joseph Choi), Clover Labs · Superwall · 4:52

24. Cap engagement in a kids AI product to protect long-term interest

    Dojo Sparks teaches reading using the science of reading, adapts difficulty with AI and listens to the child speak. The team kept sessions to about 15 minutes a day rather than unlimited screen time, guided by the principle of always doing what is in the long-term best interest of kids.

    Early results show kids gaining two grade levels in reading, about two years of progress in four or five months at 15 minutes a day.

    Sam Chaudhary, ClassDojo · Subversive · 35:50

25. Sequoia sold winners too early, and Leone says every venture investor does the same

    Leone says the best companies keep compounding for 20 years after the IPO, and winners always surprise you on the upside. Sequoia sold its Cisco stake for a total gain of 90 million while owning a quarter of the company, and Leone pushed distributions at ServiceNow that left a 10x on the table. His lesson is that if you sense an unbounded market, you should hardly ever sell.

    Sequoia owned about 20% of Nvidia, 10% of Google and a quarter of Cisco, whose value Leone puts around half a trillion dollars at its height.

    Doug Leone, Sequoia Capital, Sequoia Capital · David Senra · 24:12

26. Walt Disney lost Oswald because he owned no IP and no employment contracts

    In early 1928 Charles Mintz, the Oswald distributor, secretly signed nearly all of Walt's animators and told him to accept $500 less per cartoon. Walt had no rights to Oswald, no contracts with his staff and no customer contract, so the studio was worth effectively zero. He and Roy never forgot it, and it shaped how Disney protected its characters and talent afterward.

    Walt had to finish the Oswald contract and pay the departing animators, then built Mickey Mouse as IP he owned. Disney only got Oswald back in 2006.

    Walt Disney and Roy Disney, The Walt Disney Company, The Walt Disney Company · Acquired · 38:30

27. Larry Ellison nearly destroyed Oracle in 1991 by letting salespeople run without controls

    Sales teams booked multi-year deals up front and recognized the revenue immediately, even for software not delivered or paid for. Salespeople wrote their own contracts, gave huge discounts to customers who did not need the software for years, sold to firms that could not pay, and even did barter deals such as software for jets. Ellison admits that keeping legal and accounting as separate control functions would have avoided most of it.

    Oracle had to restate earnings and its stock fell by over 80%.

    Larry Ellison, Oracle, Oracle · Founders · 30:29

28. Track revenue metrics, not just trial conversion, when judging paywall tests

    VEED looked at trial conversion, paid conversion, MRR and ARPU for the same test. The gain in MRR (33%) was far larger than the gain in paid conversion (3%), which shows that a paywall change can affect revenue per user and plan mix beyond simple conversion.

    3% higher conversion to paid but 33% higher MRR and 17% higher ARPU.

    VEED · abtest.design

29. Aim for a 30 to 40% hook rate on video ads

    Track hook rate and hold rate as custom metrics in Meta Ads. Hook rate of 30 to 40% is the target. Hold rate can be measured as 15-second watchers over impressions, or as 15-second watchers over 3-second watchers, so the denominator changes the number.

    Ivy from Resto Me used hook rate 30 to 40% and a similar hold rate range as her benchmarks.

    Steve P. Young, App Masters · App Masters · 3:33

30. Price a two-year plan at about 30% off the annual rate

    Savvy Navvy sells an annual plan at $129 and a two-year plan at $183, which is about $75 off. This gives customers a clear saving while still pulling cash forward.

    $129 per year versus $183 for two years, roughly a 30% discount.

    Jelte Liebrand, Savvy Navvy · Sub Club · 58:59

31. Track logo churn alongside net revenue retention, not NRR alone

    NRR can look healthy while customer counts fall too fast, leaving too few customers to upgrade. Percentage losses also need larger percentage gains to recover: a 20% loss needs a 25% gain. Use both metrics to stay honest about what cohorts are doing.

    Of over 100 public SaaS companies, about two have NRR under 100%, and the median NRR at IPO is 119%.

    Jason Cohen, WP Engine · Lenny's Podcast · 54:00

32. Make the first onboarding button ask for health data sync

    Skip the intro slides and put the data-sync call to action on screen one, with a curiosity hook like 'Your Apple Watch knows what makes you tired and stressed.' Do not pitch the full solution there. Each time the team tried a prelude, it did worse.

    Over 90% of users hook up their health data, which she calls remarkable. She admits it has not been tested thoroughly.

    Asia Paloni, Welltory · Superwall · 27:28

33. Pick creators by their worst posts, not their best

    Look for creators who consistently get a high minimum view count, a baseline like 100k views per video, rather than one-hit wonders. Paid ads perform worse than average content, so judge a creator by how their worst video performs. If they can't hit the number on their worst post, it usually won't work out.

    Cal AI looks for 100k views per video consistently.

    Daniel Heintzman (host: Phil Carter), Cal AI · Subversive · 12:22

34. Dana White learned live production through trial and error on five events a year

    After buying the UFC with no production experience, White ran only five events a year at first, which gave time to learn between shows. He tried WWE-style pyro and fireworks, then found the sweet spot of a simple, technology-driven show. He still tries to be first or among the first to adopt new technology.

    The UFC now runs 43 or 44 events a year.

    Dana White, UFC, UFC · David Senra · 2:24

35. Woodruff buried a blind taste test showing people preferred Pepsi, and it came back 20 years later

    In the late 1950s McCann Erickson ran a blind taste test that found people preferred Pepsi. Woodruff ordered it never shared and never repeated. In 1975 a Pepsi bottler's agency rediscovered the same result.

    Pepsi's marketing under John Scully took the Pepsi Challenge national through local bottlers; Pepsi passed Coke in bottled share and outspent it in advertising in 1977.

    Coca-Cola's founders and leaders (Pemberton, Candler, Woodruff, Goizueta), The Coca-Cola Company · Acquired · 1:57:00

36. Morita learned that unique technology alone does not sell, and found buyers who needed a tool

    Morita and Ibuka assumed a good product would bring orders, but their first tape recorder flopped when pitched to ordinary people as an expensive toy. Morita realized he would have to be the company's merchandiser and find buyers with a real problem. Japan had a shortage of court stenographers, so he demonstrated the recorder to the Supreme Court.

    Sony sold 20 machines to the Japanese Supreme Court right away.

    Akio Morita, Sony, Sony · Founders · 34:25

37. Explain the trial so users opt in to push notifications

    When the trial flow explains that a reminder will arrive before the charge, users have a clear reason to allow push notifications. Tie the permission request to a benefit the user cares about.

    Push notification opt-ins rose from 6% to 74%.

    Blinkist · abtest.design

38. Require a minimum view count before paying the base rate

    Tie the base payment to a floor so you do not pay for posts that go nowhere. She uses 500 views, though other companies set this differently.

    She usually requires a minimum of 500 views to earn the base rate.

    Nicole Cheung, Glam Up · App Masters · 0:12

39. Test ad creative in cheap markets with similar conversion rates

    Testing around 500 ads a week at US CPMs will bankrupt you. Run the same tests in countries that convert similarly, such as Indonesia, then bring winners back to the US. Localize with AI in a few hours.

    The same test results cost $10 a day instead of $1,000 a day.

    Yuliya Lennox, Solid Starts · Sub Club · 28:57

40. Point an agent at leads your reps skipped

    Reps skip leads when commission is too small or they are busy. Train an agent on the same prompt and send it to those abandoned inbound leads. Anyone with a few thousand leads or website visitors has enough data to benefit.

    SaaStr's Agentforce agent got a 70% response rate from leads that sales had decided were not worth their time.

    Jason Lemkin, SaaStr · Lenny's Podcast · 38:30